98% win rate on 330 trades but somehow still underwater $39. 0x12a23e is a crypto bot grinding low-volatility Bitcoin prediction markets at scale. Bot executes 121 trades per day across 2,053 different markets, averaging $29 per position. The math is brutal: despite winning 98% of bets, the bot sits at -0.04% ROI with only $56 left in its portfolio. This is what algorithmic death by a thousand cuts looks like. The edge is pure volume and execution speed. Bot can hit markets milliseconds after they drop, capturing tiny inefficiencies that humans can't see. Win rate proves the model works — 98% accuracy on binary predictions isn't luck. But those wins average maybe $1 each while losses hit harder. Worst single loss sits at -$6.63 versus best win of $21.72. The bot made money on Bitcoin Up or Down - February 6, 8:15PM-8:30PM ET but couldn't escape the broader bleed. 280 open positions means the bot is still hunting, still believing the model works. Here's the contrarian read: this is what happens when you optimize for win rate instead of expectancy. A 98% win rate that loses money is worse than useless — it's a distraction. The bot trades like someone who's mastered checkers but entered chess. Low risk classification makes sense given position sizes, but low risk doesn't mean profitable. Current portfolio value of $56.23 after $102K in volume suggests the bot is down to final attempts at recovery. Either the model recalibrates or this becomes a cautionary tale about automation without edge.